OPINION
PHILLIPS, Circuit Judge:The question is the entitlement of certain government employees, appellees in this appeal, to an award of attorney fees as prevailing parties under the Equal Access to Justice Act, 28 U.S.C. § 2412 (the EAJA), following the Government’s unsuccessful attempt to recover from them a civil penalty for allegedly receiving from their former private employer compensation in violation of 18 U.S.C. § 209(a). Because the former employer is legally obligated to indemnify the employees for their attorney fees in the event of their successful defense of the Government action, we conclude that they did not “incur” the fee expenses, hence are not eligible for a fee award under the EAJA. Alternatively, we conclude that in any event the Government’s position in the underlying litigation was “substantially justified,” so that on that basis as well the employees are not entitled to a fee award under the EAJA. Accordingly we reverse the order of the district court awarding fees to the employees.
*1163I
In 1986, the United States brought a civil penalty action against The Boeing Company, Inc. (Boeing), and appellees, five former Boeing employees, alleging that the lump sum payments appellees received as “severance payments” from Boeing upon leaving their employment with that company, but before accepting government appointments, violated 18 U.S.C. § 209(a), a criminal statute prohibiting payment of any government employee’s salary by anyone other than the Government or receipt of supplemental salary by a government employee from an outside source. Boeing advanced the attorney fees and expenses of four of the five appellees pursuant to agreements obligating these appellees to repay Boeing all amounts advanced “unless it shall ultimately be determined pursuant to Section 145 of the Delaware Corporation Law and Section 4 of Article VII of the ByLaws of the Company” that appellees are entitled to indemnification by Boeing.
The district court held that the payments did not violate 18 U.S.C. § 209(a); a divided panel of this court reversed; and the Supreme Court, on writ of certiorari, reversed this court, holding that the payments did not violate section 209(a) because the statute does not apply to payments made to or received by persons who, like appellees, were not government employees at the time the payments were made and received. United States v. Boeing Co., 653 F.Supp. 1381 (E.D.Va.1987), aff'd in part and rev’d in part, 845 F.2d 476 (4th Cir.1988), rev’d sub nom. Crandon v. United States, 494 U.S. 152, 110 S.Ct. 997, 108 L.Ed.2d 132 (1990).
Appellees then applied to the district court for an award of attorney fees and costs of defending the action under the EAJA, which, in pertinent part, provides that a court
shall award to a prevailing party other than the United States fees and other expenses ... incurred by that party in any civil action (other than cases sounding in tort) ... unless the court finds that the position of the United States was substantially justified or that special circumstances make an award unjust.
28 U.S.C. § 2412(d)(1)(A).
The Government opposed appellees’ EAJA claim by arguing that because controlling Delaware law obligated Boeing to indemnify appellees, only Boeing, and not the appellees, actually “incurred” the fees in this action.1 The Government also argued that even if appellees were held to have incurred the fees and expenses in this matter, the Government had been “substantially justified” in bringing and prosecuting the underlying action against appel-lees. Kejecting both of the Government’s arguments, the district court granted ap-pellees’ application for an EAJA award. United States v. Boeing Co., 747 F.Supp. 319 (E.D.Va.1990).
This appeal by the Government followed.
II
The Government challenges both grounds upon which the district court found the appellees entitled to a fee award: (1) that appellees incurred the fees and costs at issue, and are therefore eligible to recover them under the EAJA; and (2) that the Government’s position in the underlying merits litigation was not substantially justified. We take these in order.
A
The Government argues that because Del.Code Ann. tit. 8, § 145(c) unconditionally required Boeing to indemnify appellees for all of their attorney fees and costs, appellees did not “incur” those expenses, an essential condition to entitlement under the EAJA. Section 145(c) provides:
To the extent that a director, officer, employee or agent of a corporation has been successful on the merits or otherwise in defense of any action, suit or proceeding referred to in subsections (a) *1164and (b) of this section, or in defense of any claim, issue or matter therein, he shall be indemnified against expenses (including attorneys’ fees) actually and reasonably incurred by him in connection therewith.
We agree with the Government.
The underlying merits litigation was of a type referred to in subsection (a) of section 145 of the Delaware Code, so that its subsection (c) applies.2 Appellees indisputably were successful on the merits. Under these circumstances, section 145(c) plainly directs that appellees “shall be indemnified” by Boeing and we therefore hold that Delaware law requires Boeing to indemnify appellees for the attorney fees and costs of defending the underlying action.
Appellees contend, however, that even if Delaware law gave them an unconditional right to be indemnified, this does not resolve the question whether they nevertheless should be deemed to have “incurred” the expenses for EAJA purposes. They argue that they should be deemed to have incurred them because having paid them, Boeing’s indemnification (or, more properly, declination to demand return of its advances) can only be counted at this point “a mere possibility.” The district court accepted this argument, Boeing, 747 F.Supp. at 321. We disagree.
To hold that a prevailing party with an unconditional legal right to indemnification of its attorney fees by a manifestly solvent third party might nevertheless qualify for an EAJA award because indemnification had not yet occurred is unacceptable for several reasons. In the first place, it necessarily would imply an assumption by a court that the indemnitor either would default deliberately on its legal obligation, or, even more questionably, would conspire with the indemnitee to alter the intended operation of the EAJA by jointly disregarding the legal right and duty.
Aside from the unseemliness of such a judicial assumption, such a holding would not in fact serve a principal purpose of the EAJA: to avoid the deterring effect which liability for attorney fees might have on parties’ willingness and ability to litigate meritorious civil claims or defenses against the Government. The EAJA provides for feeshifting precisely to avoid this result. See SEC v. Comserv Corp., 908 F.2d 1407, 1415 (8th Cir.1990). Consequently, in any situation in which the eligibility of a particular prevailing party for an EAJA award is in issue, it is appropriate to inquire whether that party would, as a practical matter, have been deterred from litigating had it been known that a fee-shifting award was not available upon a successful conclusion. If that question is asked here, it is obvious that appellees, all but one of whom were funded by an advance which by contract they need not refund if they prevailed in litigation, would not have been deterred had the EAJA not then existed. This is the critical concern underlying the EAJA precondition that a fee claimant shall have “incurred” the expense.
Accordingly, we hold that, to effectuate the purposes of the EAJA, a claimant with a legally enforceable right to full indemnification of attorney fees from a solvent third party cannot be deemed to have incurred that expense for purposes of the EAJA, hence is not eligible for an award of fees under that Act.
B
The Government also contends that even if appellees should be deemed to have “incurred” these expenses, thereby qualifying *1165for a fee award if the other requirements of the EAJA were met, the district court also erred in concluding that the Government’s position was not substantially justified. We agree.
Whether for purposes of the EAJA the Government’s “position” in particular litigation is “substantially justified” has proved to be an issue of considerable conceptual and practical difficulty, given the open-endedness of the statutory language and, no doubt, the delicacy of the question. It had given the lower federal courts enough trouble that in 1987 the Supreme Court undertook to clarify the matter — so far as was possible in light of the statutory text. In Pierce v. Underwood, 487 U.S. 552, 108 S.Ct. 2541, 101 L.Ed.2d 490 (1988), the Court, wrestling essentially with the two possible (almost opposite) meanings of “substantially” in this context — “considerable” (to a high degree) or “in the main” (to a reasonable degree) — came down with “in the main,” i.e., “to a degree that could satisfy a reasonable person.” Id. at 565. That comes out, said the Court, as “no different from the ‘reasonable basis in law and fact’ formulation adopted by ... the vast majority of Courts of Appeals that have addressed this issue,” including this court, see Anderson v. Heckler, 756 F.2d 1011, 1013 (4th Cir.1985).
In the context of this case, the question of substantial justification is therefore whether the Government’s “position” — that the severance payments to the Boeing employees were prohibited by 18 U.S.C. § 209(a) — was one that had a reasonable basis in law and fact, i.e., was such as to satisfy a reasonable person. In practical terms, this comes to whether the Government attorneys who decided to bring the civil penalty action made a reasonable legal decision that § 209(a) covered the payments.
The district court held that this decision was not a reasonable legal one, that the position taken on its basis was not substantially justified. The court’s conclusion was rested almost entirely on the fact that the legal position taken by the Government had in the end been found legally wrong by a unanimous Supreme Court (not to say by the district court and a dissenting court of appeals judge on the way to the Supreme Court). 747 F.Supp. at 322.
We are required to review this decision of the district court under an abuse of discretion standard. Pierce, 487 U.S. at 557-63, 108 S.Ct. at 2546-49. Though recognizing that such a decision may involve both factual, legal, and discretionary components, the Pierce Court concluded that, mainly for concerns of judicial administration, a unitary standard that accords the deference traditionally associated with abuse of discretion review was appropriate for all aspects of this decision, including even those that could be characterized as purely legal. Id.
Some of the practical reasons thought in . Pierce to justify deferential review even of purely legal aspects of substantial justification decisions are not present where, as here, the merits issue has by now been finally and definitively resolved. In such a situation, there is not the general danger emphasized in Pierce that intensive de novo review of the legal aspects of an EAJA decision might indirectly compromise a still pending review of the closely-related merits issues, or encourage needless merits appeals by the Government to avoid the risk of indirect “affirmances” (on fee-claimants’ EAJA appeals) of Government losses on the merits in the district courts. See id. at 561, 108 S.Ct. at 2548.
Furthermore, since Pierce was decided, the Court has indicated that in the comparable use of a unitary abuse of discretion standard in reviewing Rule 11 sanction decisions, flat errors of law must be considered abuses of discretion. Cooler & Gell v. Hartmarx Corp., 496 U.S. 384, -, 110 S.Ct. 2447, 2459, 110 L.Ed.2d 359 (1990).
Notwithstanding these intimations that pure legal error must yet on occasion be freely reviewable under these administratively-driven deferential review regimes, Pierce still stands powerfully against such a view. See Pierce, 487 U.S. at 563, 108 S.Ct. at 2549 (conceding apparent need in some cases for more stringent review stan*1166dard to avoid injustice, but nevertheless concluding that “the generality rather than the exception must form the basis for our rule [of deference]”). We therefore consider that even with respect to misapprehensions of controlling law, our review must be deferential in obedience to Pierce.
There remains the question—how deferential? Pierce of course wisely did not attempt to formulate the degree in terms more precise than the traditional abuse-of-discretion formulation it adopted for across-the-board application. But Pierce obviously contemplated a degree of deference considerably short of a simple, accept-on-faith, rubber-stamping of district court decisions on this issue. Pierce’s own careful, detailed review to determine whether the primary EAJA decision in that case reflected an “abuse of discretion” demonstrates that the deference commanded is a suitably informed deference. See Pierce, 487 U.S. at 568-71, 108 S.Ct. at 2551-53 (affirming court of appeals holding of noa-buse in district court’s determination that Government’s position was not substantially justified).3
In the course of its review, Pierce also gave some guidance on the way to conduct principled review under this standard. It first assessed the district court decision by looking at the available “objective indicia” of the strength of the Government’s position: in that case, the terms of the settlement agreement that ended the underlying litigation, the stage at which the merits were thereby decided, and the views of other courts on the strength, hence reasonableness, of the Government’s position. Essentially finding these a washout, certainly not sufficiently favorable to the Government to find on their basis alone an abuse of discretion in the district court’s decision, the Court turned to an independent assessment of the merits of the Government’s position. Finding that the opposing arguments advanced by the parties on the unadjudicated merits did not “command” a conclusion that the Government’s position was substantially justified, the Court concluded that the district court had not abused its discretion in its determination to the contrary. See id. at 568-71, 108 S.Ct. at 2551-53.
■We can follow the general review process Pierce charted, but the available indicia inevitably are not at all the same. This case was litigated all the way to the Supreme Court, not settled before trial. No other court has ruled on the merits of the issue in other cases. Here the obvious sources for an independent assessment of the strength of the Government’s merits position are the merits, decisions and opinions of the district court, the court of appeals, and the Supreme Court, rather than mere opposing contentions of counsel as to the merits.
As did the Pierce Court, we look first to the available “objective indicia.” Here the merits were finally decided against the Government only at the Supreme Court level. We think this at best a neutral factor. Although the fact that the claim survived through the court of appeals rather than collapsing or settling at an earlier stage points generally in the direction of strength, it obviously could not suffice, standing alone, to command a finding of substantial justification. Completely unfounded claims sometimes, for a variety of reasons, survive beyond their just desserts.
The fact that no other court has ruled on the merits—none having had a *1167chance — is advanced by the appellees as an indication of nonjustification. The fact that although § 209(a) had been on the books for 27 years, this was the first action ever brought by the Government on comparable facts indicates, they contend, the weakness of the claim, hence non-justification for its prosecution. The district court thought this persuasive. 747 F.Supp. at 322. We disagree.
We see this as a completely neutral factor in assessing the strength of the Government’s position. There are any number of practical reasons other than awareness of a claim’s intrinsic weakness why its attempted application by Government enforcers to a particular type of conduct may be delayed. Though this was a claim for civil penalty rather than a criminal prosecution, the Government’s interests here were essentially those that inform prosecutorial discretion in criminal matters. That discretion must and does involve the targeting of particular conduct at particular times in response to many legitimate factors having nothing to do with the general legal strength of the claim.
When we turn to an independent assessment of the strength of — the justification for — the Government’s position, the most immediate direct indicator is that it has been rejected on the merits by the Supreme Court. As earlier noted, this was the principal basis upon which the district court found it not substantially justified for EAJA purposes.
Although the impulse to equate ultimate judicial rejection of the Government’s merits position — at whatever level — with its lack of substantial justification is understandable, the courts perforce have rejected it as inappropriate, for to do so, “would virtually eliminate the ‘substantially justified’ standard from the statute.” Broad Avenue Laundry & Tailoring v. United States, 693 F.2d 1387, 1391-92 (Fed.Cir.1982). Or, as Judge Mumaghan of this court once put it, it would be “a war with life’s realities to reason that the position of every loser in a lawsuit upon final conclusion was unjustified.” Evans v. Sullivan, 928 F.2d 109, 110 (4th Cir.1991).
Equally unacceptable is the opposite impulse: to find in an intermediate judicial determination of merit in the Government’s position proof that the position was at least one that “could satisfy a reasonable person,” i.e., the one or more presumably reasonable Article III judges who at some stage of the litigation found merit in it. That too, despite the possible incongruity of not so concluding, cannot be allowed to determine the matter. To do so would be equally at “war with life’s realities” in the world of judges and judging and the legal process. As a practical matter, the substantial justification issue cannot be transformed into an up-or-down judgment on the relative reasoning powers of Article III judges who may have disagreed on the merits of a Government litigation position.
In sum, merits decisions in a litigation, whether intermediate or final, cannot, standing alone, determine the substantial justification issue. But of course they— and more critically their rationales — are the most powerful available indicators of the strength, hence reasonableness, of the ultimately rejected position. As such, they obviously must be taken into account both by a district court in deciding whether the Government’s position,. though ultimately rejected on the merits, was substantially justified, and by a court of appeals in later reviewing that decision for abuse of discretion.
In this case, the district court accorded no weight to the cogency of the reasoning by which two circuit judges of this court had concluded that the Government’s position was not only justified, but was legally correct.4 That reasoning was, essentially, *1168that the critical language of § 209(a) did not “[o]n its face .,. require that payment occur while the [recipient] was a government employee”; that in total statutory context, the intended legislative meaning on this critical point was ambiguous, requiring resort to legislative history; and that this history pointed to an intention to cover the type pre-government employment payments at issue. 845 F.2d at 479-80.
Indirectly alluding to the court of appeals’ reasoning—which of course was relied upon by the Government in attempting to establish the substantial justification for its position—the district court considered it to have been flatly rejected by the Supreme Court. As to whether the critical language was, in the first place, ambiguous on its face—as the court of appeals had held—the district court thought that the Supreme Court had “unequivocally concluded that the ‘text of [Section] 209(a) ... indicates that employment status is an element of the offense.’ ” 747 F.Supp. at 322 (quoting Crandon, 110 S.Ct. at 1002). On this basis, the district court concluded that “[i]t was, therefore, unreasonable for the government to rely on a nonexistent ambiguity in the statute.” Id.
Furthermore, opined the district court, even if the statute were ambiguous the authority construing Section 209, as well as the pertinent legislative history, provides no support for the government’s position. As the Supreme Court found, the government’s attempt to apply the statute to persons not serving as federal employees at the time of the payment was at odds with the language and clear congressional intent of Section 209.
Id. (citing Crandon, 110 S.Ct. at 1003-06).
In sum, the district court read the Supreme Court’s rationale for reversal of the court of appeals’ decision as so thorough and ready a rejection of that court’s statutory interpretation as to compel the conclusion that the Government’s position thereby upheld was not substantially justified.
With all respect, and with the deference we think Pierce commands, we think the district court so seriously underestimated the Supreme Court’s recognition of the difficulty of the interpretive issue, hence of the reasonableness of the Government’s position, that its conclusion cannot be allowed to stand. At odds with the district court’s reading of the Supreme Court’s decision, we think that decision, fairly read, demonstrates that the interpretive issue was sufficiently difficult and arguable that the Government was completely justified in seeking to apply § 209(a) to the conduct at issue. Our assessment rests both on the general indications of difficulty reflected in the Supreme Court’s decision and on a particular feature of the underlying litigation that—as the Supreme Court decision implicitly recognized—warrants special deference to the Government’s litigation position in this type case.
First off, the Supreme Court decision was not, as it might have been had the interpretive issue been considered an open- and-shut one, by summary reversal. Cf., e.g., Palmer v. BRG of Georgia, Inc., — U.S. -, 111 S.Ct. 401, 112 L.Ed.2d 349 (1990) (per curiam) (summary reversal of court of appeals decision respecting reach of § 1 of the Sherman Act). Instead it prompted two separate, fairly lengthy, opinions, each of which demonstrates a perception by the author that the issue was one whose resolution required significant intellectual effort, careful analysis of relevant interpretive sources, and sophisticated applications of traditional canons of statutory construction. See Crandon, 110 S.Ct. at 999-1007 (Stevens, J., majority opinion); *1169Id. at 1007-15 (Scalia, J., with O’Connor and Kennedy, JJ., concurring in the judgment).
Although the two opinions agree in the end that § 209(a) could not properly be interpreted to cover the pre-Government employment lump-sum severance payments there in issue, they differ on the proper interpretive route to that end. Critically, both seem to agree that the most directly applicable statutory text (that of the section’s “receiving” prohibition), does not, as a matter of plain, unaided meaning, dictate the answer. See id. at 1002 (majority opinion) (“neither [of § 209(a)’s prohibitions] directly specifies when a payment must be made or received”); id. at 1007-08 (concurring opinion) (opining that neither prohibition, literally read, excludes the possibility that some types of payment made before or after Government employment might be covered). Both then find, with the aid of a wider contextual analysis of the specifically applicable prohibition, legislative history, and indications of general statutory policy, meanings that exclude from coverage the severance payments there at issue, but the meanings so found are different. The majority finds a congressional intent to prohibit only payments received during the period of Government employment, id. at 1003-06; the concurrence, an intent to prohibit only those payments, whenever made, that in some form could be considered periodic contributions to or supplementations of Government salaries, id. at 1008-10.
We are satisfied that any legally trained reader of the two Crandon opinions would see in them a Court concerned with a serious issue of statutory interpretation.
Passing these general indications of the interpretive difficulty posed by § 209(a), the most critical specific aspect of the Crandon Court’s statutory analysis is its reliance on the rule of lenity in resolving the interpretive issue against the Government’s position. Pointing out at the beginning of its analysis that “the governing standard is set forth in a criminal statute,” the Court noted that this made it “appropriate to apply the rule of lenity in resolving any ambiguity in the ambit of the statute’s coverage.” Id. at 1001. And the Court concluded its analysis of statutory text, legislative history, and statutory design and policy, see id. at 1001-06, by reiterating its reliance on this interpretive rule in construing the reach of this criminal statute.
Finally, as we have already observed, we are construing a criminal statute and are therefore bound to consider application of the rule of lenity. To the extent that any ambiguity over the temporal scope of § 209(a) remains, it should be resolved in [the employees’] favor unless and until Congress plainly states that we have misconstrued its intent.
Id. at 1007; see also id. at 1008, 1010 (concurring opinion) (drawing on rule of lenity).
Implicit in this of course is a perception by the Court that the statute had enough of ambiguity about it to require invocation of the rule of lenity, for that is the only occasion for its invocation. See Liparota v. United States, 471 U.S. 419, 427, 105 S.Ct. 2084, 2089, 85 L.Ed.2d 434 (1985). This implication is reinforced by various specific allusions in the two opinions to the ambiguity of the § 209(a) text. See Crandon, 110 S.Ct. at 1005 (majority opinion) (“appropriate, in a case that raises questions about the scope of the prohibition, to identify the specific policies that the provision serves”); id. at 1006 (majority opinion) (noting that interpretation finding coverage under § 209(a) serves one of the two principal statutory policies, disserves the other); id. at 1010 (concurring opinion) (conceding that proffered interpretation “is no more successful than the Government’s in giving effect to all the language of the section,” and concluding on the particular interpretive problem that “at most we have an ambiguity”).
We therefore think that the district court’s perception that Crandon’s ultimate unanimous rejection of the Government’s position necessarily demonstrated that ambiguity justifying the Government’s position was “nonexistent,” is flatly belied by a careful reading of the Crandon Court’s two opinions. Instead, in its recognition of *1170the need to draw on the rule of lenity to resolve the interpretive issue, the Court implicitly recognized that the Government’s position, taking a broader view of the statute’s reach than lenity allowed the Court to take, might well be “justified,” given the different roles of the executive and judicial branches in enforcing the criminal laws. For as Justice Scalia pointed out in his concurring opinion, the Government’s enforcement role requires that as between the opposing risks of taking too narrow or too broad a view of “what it may prosecute,” it must in prudence choose the broad view, knowing that judicial review (including lenity in the end) stands guard against error in that choice, whereas an error in the opposite direction is not likely ever to be corrected. Id. at 1011.5 In effect, this reflects a judicial perception that, within bounds of essential prosecutorial fairness, Government prosecutors have an obligation to establish the outer limits of a criminal statute's ambiguously-defined reach by making just such choices when occasion requires.
Here, despite the deference we must accord even those district court EAJA decisions that we may think legally “wrong,” we think this one sufficiently wrong, with sufficiently dangerous implications for the Government's prosecutorial function, that we must consider it, under Pierce’s standard, an “abuse of discretion.” It reflects, as we have attempted to demonstrate, an insufficiently careful assessment of the true import of the Crandon opinions as they bear upon the strength, hence justification for, the Government’s litigation position. And it necessarily implies a more cramped view of the latitude that must be accorded Government interpretations of the reach of untested, ambiguously-drawn criminal statutes than is acceptable.6
We therefore hold that, contrary to the district court’s determination, the opposite conclusion, that the Government’s litigation position in the underlying litigation was substantially justified, is “commanded,” see Pierce, 487 U.S. at 570, 108 S.Ct. at 2552-53, by the most relevant indicators on that issue.
REVERSED.
. Boeing does not qualify for recovery of fees under the EAJA. A corporation is eligible for an EAJA award only if its net worth did not exceed $7,000,000 and it had no more than 500 employees at the time the action was filed. 28 U.S.C. § 2412(d)(2)(B).
. Each appellee was a party to the underlying action “by reason of the fact that he is or was a director, officer, employee or agent of the corporation." Del.Code Ann. tit. 8, § 145(a). In their brief, appellees conjecture that Boeing might decline to indemnify them, taking the position that appellees were sued by the Government because of their employment with the United States, not because of their employment with Boeing. However, it was only "by reason of the fact” that appellees were employees of Boeing and received payments from Boeing before going to work for the Government that they were parties to the action brought against them (and Boeing) by the Government. Any effort by Boeing to avoid liability for indemnification on that basis (a basis, it should be noted, which appellees carefully decline to suggest as more than a theoretical possibility) would not succeed.
. Although one of the avowed aims of the adoption of an across-the-board abuse of discretion review standard is the avoidance of duplicative judicial efforts, see Pierce, 487 U.S. at 560-61, 108 S.Ct. at 2547-48, it is inescapable that honest, principled review under the traditional deferential standards may yet require considerable appellate court effort. In order to declare that discretion has been abused, or clear error committed in finding facts, a substantial canvassing of the relevant materials may be required in any case of significant complexity. See Anderson v. City of Bessemer City, 470 U.S. 564, 581, 105 S.Ct. 1504, 1515, 84 L.Ed.2d 518 (1985) (Powell, J., concurring) (cautioning that Court’s deliberate heightening of deference to be accorded in clear error review of factfindings should not be read to relieve appellate courts of the "type of burdensome review that may be appropriate in some cases”). Review of "discretionary” EAJA decisions can be expected frequently to require the same sort of "burdensome” review.
. For the reasons just noted, the court was of course right in holding that the court of appeals decision could not "in itself’ establish substantial justification. 747 F.Supp. at 322. But this does not mean, on the other hand, that the rationale of an ultimately rejected intermediate decision is, on that account alone, entitled to no weight. That may or may not be the case, depending upon the cogency of rationale for that decision, as that, in turn, is mainly indicated by the rationale for its later reversal. The point is that the mere cast of intermediate and *1168final merits decisions, or simple head-counting of judges on different sides of the merits issue, cannot be made decisive. An independent assessment of the substantial justification for the Government position requires a deeper and more difficult inquiry.
Interestingly, the district court apparently thought it necessary to discount the significance of the court of appeals’ intermediate reversal of the former’s ultimately vindicated merits decision. This the court did by pointing-erroneously—to Pierce's requirement of deference to district court decisions on the substantial justification issue itself. Id. As noted above, the proper view would be that the court of appeals’ reversal—standing alone—was of no significance, only the cogency of its rationale was.
. Justice Scalia’s point was not of course being made with reference to the EAJA substantial justification issue, but in addressing the merits. He was explaining why the Government’s consistent pattern of interpreting § 209(a) to cover such payments as those in issue was due no judicial deference. To accord deference, he pointed out, would in effect be to convert the judicial rule of lenity into one of judicial "severity” in construing the reach of a criminal statute, since the proper prosecutorial rule is not lenity, but severity, i.e. broad interpretation. Though made in reference to the merits issue, it perfectly anticipates the substantial justification issue.
. If, as we suggest, special latitude for Government litigation positions respecting the reach of criminal statutes may be due in assessing their justification for EAJA purposes, the special case thereby created is an exceedingly narrow one. It is limited to those relatively rare occasions when the Government has sought, as here, to recover a civil penalty for violation of a criminal statute. The EAJA does not, of course, apply to criminal prosecutions.